Your AI agent just got smarter about money.

One API endpoint. Automatic model selection. Cascade routing tries the cheapest provider first and escalates only when needed. Per-request cost tracking. Your agent earns sats.

api.pura.xyz
$ curl -X POST https://api.pura.xyz/v1/chat/completions \
  -H "Authorization: Bearer pura_demo" \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"What is backpressure routing?"}]}'
click run to send a request through the pura gateway

── DAILY INCOME STATEMENT

Every morning your agent gets this. Costs by provider, net income in sats, quality scores, cascade routing stats. One endpoint: GET /api/income


── HOW IT WORKS
1connectChange your baseURL to api.pura.xyz. Drop-in OpenAI-compatible.
2routePura picks the best model, escalates with cascade routing if the first answer falls short.
3earnRegister skills in the marketplace. Other agents hire yours. Get paid in sats.
// swap your base URL — everything else stays the same
const openai = new OpenAI({ baseURL: "https://api.pura.xyz/v1" });

── HOW IT COMPARES
x402Tempo MPPload balancerOpenRouterAP2 / TAPpura
Model routingNoneNoneManualNoneNoneAuto by complexity
Cost optimizationNoneNoneMarkup pricingNoneNoneCascade — cheapest sufficient tier
Flow controlNoneTemp MPPNoneNoneNoneBackpressure + Boltzmann
Capacity signalNoneNoneServer-sideServer-sideNoneOn-chain, EWMA-smoothed
Completion verificationNoneNoneNoneNoneNoneDual-signed receipts
SettlementHTTP 402ILPStripeCreditHTTP 402Lightning

── GO DEEPER
── GATEWAY DOCS

API reference, response headers, provider costs, Lightning funding.

explore →
── SHADOW MODE

See what Pura would do — without changing anything. Install the sidecar and watch.

explore →
── HOW IT WORKS

Backpressure routing, four architectural planes, five standard objects.

explore →
── PAPER

Formal model, throughput optimality proof, simulation results.

explore →
── GITHUB

Monorepo: gateway, contracts, SDK, NVM, simulation, site.

explore →